The intention (for now) of this page is to define the Scope (Anwendungsbereich) for the DIN-SPEC-Project on "Wissensgraphen und Ontologien in großen Sprachmodellen"
3 types of KG+LLMs systems (Robert D. - SWC)
- Complementary solution:
KGs and LLMs complement each other to implement a use case that each would not be able to do. [I wouldn't restrict to the point that alone KG and LLM would not be able to solve the task. Probably better to say that their complementary use enhances the performance/quality/efficiency/... (significantly)]- KG-based (Semantic or Graph) RAG
- Improve LLMs by using KGs
- reduce hallucinations/garbage
- provide additional background knowledge
- ...
- Improve KGs by using LLMs
- improve structure of KGs-> extend/augment KG with nodes/entities + relations
- content -> add labels/descriptions based on other KG entities
- Context-aware translation of labels/literals between different languages
- Disambiguate KG entities
- LLMs for entity matching in different KGs
- LLMs for ontology engineering tasks (in ontology design, ontology mapping/alignment,...)
- ...
Based on above version (Harald, Sven, Heike - FIZ Karlsruhe)
- Definition of vocabulary used
Only focus on language (LLMs) vs multi-modal (language + images, audio, etc)?
- Complementary solution
- Downstream tasks
- Question answering
- Fact checking
- Fake news detection
- Explainability
- Downstream tasks
- Improve LLMs by using KGs
- KG-enhanced LLM training
- Integrating KGs into training objective
- Integrating KGs into LLM inputs (verbalize KG for LLM training)
- Integrating KGs by fusion modules
- Retrieval-augmented Generation (RAG)
- KG-guided retrieval mechanisms (Daniel B. (FSTI))
- Hybrid retrieval combining KGs and dense vectors (Daniel B. (FSTI))
- KG-enhanced reranking of retrieved information (Daniel B. (FSTI))
- KG-enhanced LLM interpretability
- KGs for LLM probing
- KG-based analysis of attention patterns (Daniel B. (FSTI))
- Measuring KG alignment in LLM representations (Daniel B. (FSTI))
- KG-guided explanation generation (Daniel B. (FSTI))
- KG-based fact-checking and verification (Daniel B. (FSTI))
- KGs for LLM probing
- KG-enhanced LLM inference / reasoning
- KG-guided multi-hop reasoning (Daniel B. (FSTI))
- Integrating symbolic reasoning with LLMs using KGs (Daniel B. (FSTI))
- KG-based consistency checking in LLM outputs (Daniel B. (FSTI))
- KGs for LLM analysis
- Using KGs to evaluate LLM knowledge coverage (Daniel B. (FSTI))
- Analyzing LLM biases through KG comparisons (Daniel B. (FSTI))
- KG-enhanced LLM training
- Improve KGs by using LLMs
- Assertional knowledge engineering
- Information Extraction
- KG completion (A-Box)
- Link prediction
- Relation prediction
- Fact checking / Triple testing
- Literal completion (labels/comments/descriptions)
- KG completion (A-Box)
- Entity Linking (between KGs)
- Entity Disambiguation
- Information Extraction
- Terminological knowledge engineering
- Ontology Design
- Competency Question (CQ) generation
- User stories / personas generation
- Ontology learning (Automated ontology design from text)
- Ontology Evaluation
- Competency Question (CQ) generation (from given ontologies)
- CQ to SPARQL
- Ontology Mapping
- Ontology Documentation
- Class and relation descriptions/labels
- Ontology Design
- Reasoning
- Aprox/Probabilistic Reasoning via LLMs (LLM supported)
- Constraint checking (Robert D.)
- Data Repairs (→ maybe move to completion?) (Robert D.)
- Downstream tasks
- KG/Ontology embeddings
- User interface / Access
- Natural Language interface to KG
- KG to natural language (verbalization)
- Multilingual translation of literals
- Assertional knowledge engineering